Whole-Genome Expression Microarray Combined with Machine Learning to Identify Prognostic Biomarkers for High-Grade Glioma.

The aim of our study is to build a framework for a better understanding of high-grade glioma (HGG) prognostic-related biomarkers. Whole-genome gene expression microarray was performed to identify differently expressed genes between HGGs and low-grade diffuse gliomas. Several machine learning algorithms were used to filter prognostic-related genes. One hundred ninety-three HGG patients after surgical resection were selected for survival analysis. Immunohistochemistry were performed on these tumor samples to analyze IDH1 mutation status and protein expression of WEE1. qRT-PCR, western blotting, transwell assays, and scratch wound healing assays were performed to evaluate the effect of WEE1 knockdown or overexpression in HGG cells. Three prognostic-related genes (WEE1, IGF2PB3, and EMP3) were demonstrated to separate HGG patients into two different survival subgroups. The area under receiver operating characteristic curve of WEE1 was higher than that of IGF2BP3, EMP3, age, IDH status, 1p/19q status, and MGMT promoter status. WEE1 was an independent covariate compared with IDH status, age, and WHO grade. Knockdown or overexpression of WEE1 can inhibit or promote migration and invasion in U251 and U87 cell lines. WEE1, EMP3, and IGF2BP3 are reliable prognostic-related genes at the mRNA level. WEE1 is an independent prognostic biomarker in survival analysis and has potential diagnostic value for HGG patients. WEE1 can induce HGG cell migration and invasion in vitro.